mbirjax 0.6.17.1

High-performance tomographic reconstruction

Model-Based Iterative Reconstruction (MBIR) for tomographic reconstruction that is based on the JAX library. Full documentation is available at https://mbirjax.readthedocs.io .

Installing from PyPI

For CPU only:

pip install mbirjax

For CPU with a CUDA12-enabled GPU:

pip install --upgrade mbirjax[cuda12]

Installing from Source

1. Clone the repository:

   git clone [e-mail]:cabouman/mbirjax.git

2. Install the conda environment and package

   1. Option 1: Clean install using dev_scripts - We provide bash scripts that will do a clean install of MBIRJAX in a new conda environment using the commands:

      cd dev_scripts
      source clean_install_all.sh

   2. Option 2: Manual install - You can also manually install MBIRJAX from the main directory of the repository with the following commands:

      conda create --name mbirjax python=3.10
      conda activate mbirjax
      pip install -r requirements.txt
      pip install .

Optional Pixi Development Environment

For contributors who use Pixi, MBIRJAX also provides an optional reproducible development environment. This does not replace the conda installation workflow above.

For the default CPU environment on Linux or Apple Silicon macOS:

pixi run smoke
pixi run test-fast

For a CUDA-enabled Linux system:

pixi run -e cuda smoke-jax
pixi run -e cuda test-fast

Additional useful tasks include:

pixi run test
pixi run test-data
pixi run docs

Running Demo(s)

Run any of the available demo scripts with something like the following:

python demo/<demo_file>.py